# InEpoch: local sleep staging software, deployed and trained in your lab > InEpoch installs and configures local sleep-staging software in sleep laboratories, sets up the review workflow, and trains sleep technologists to use it. Study files and analysis stay on the customer's workstation; no cloud upload is required for analysis. The software proposes stages for professional review and is not an autonomous diagnosis. Canonical: https://inepoch.com/ ยท Last updated: 2026-10-11 ## How it works 1. **Every study starts as a signal.** EEG, with optional eye-movement (EOG) and muscle (EMG) channels, imported from an EDF or EDF+ recording. 2. **Scored in 30-second epochs.** Sleep staging divides the night into 30-second epochs and assigns each one a stage: Wake, N1, N2, N3 or REM. 3. **The model suggests. A person decides.** YASA proposes a stage and probabilities for every epoch. A trained reviewer confirms or corrects each one, with notes. 4. **Processed on your workstation.** Import, staging, review and export run on the workstation. No cloud upload is required for analysis. 5. **Handed over to a trained team.** Installation, workflow setup and technician training, so the software is used with care from the first study. ## Why it matters In a sleep study, the night is divided into 30-second epochs and each epoch is assigned a stage (Wake, N1, N2, N3 or REM) following rules published by the American Academy of Sleep Medicine. Visually scoring a single night can take a well-trained scorer up to two hours (Vallat & Walker, eLife 2021). The AASM has described assisted scoring of polysomnography as one of the most apparent applications of AI in sleep medicine, while setting out its limitations. ## What InEpoch provides - **Installation and configuration.** We install the desktop application on the workstations you designate, confirm where studies are stored, and test imports with your own recording setup, including channel labels, sampling rates and voltage units. - **Workflow setup and guided operation.** We map your scoring process onto the software: channel derivations, review queues, the confidence threshold that flags epochs for review, reviewer identifiers, sign-off and exports. Then we work through the first studies with your team. - **Technician training.** Hands-on sessions for technologists and reviewers, using authorized or public sample recordings: moving through epochs, reading per-epoch probabilities, correcting stages, adding notes and producing reports. - **Support agreed for your deployment.** Support channels, response expectations, update handling and any remote assistance are set out in writing for your deployment. We do not access study data unless you arrange it. ## Workflow 1. **Import a recording.** Choose an EDF or EDF+ file. The application keeps a private local copy and never changes the original. 2. **Confirm channels.** Check the actual labels, sampling rates and units, preview the signals, then explicitly confirm the EEG and any EOG and EMG derivations. 3. **Run local staging.** YASA's pretrained classifier runs on the workstation's CPU and returns a stage and stage probabilities for every 30-second epoch. 4. **Review and correct.** Step through the hypnogram and signals. Filter for low-confidence, N1, transition or unreviewed epochs. Confirm or correct each stage and add notes. The original model output and every change are kept, with undo and redo. 5. **Sign off and export.** Once every epoch has a review decision, sign off and export a per-epoch CSV and a PDF report. ## Local by design Study files and analysis stay on your workstation during the local workflow. Studies and results live in a local data folder, staging runs on the workstation's CPU, and the application's internal server accepts connections only from the same computer. Access, disk encryption, backups, and any authorised support arrangements are configured with your organisation. This website is separate from the software; its inquiry form collects contact details only. ## Training and handover - **Practical onboarding.** Short, hands-on sessions at the workstation, built around how your lab already scores, rather than a generic product tour. - **Guided review.** We review sample studies together: where the model is least certain, how corrections and notes are recorded, and when to escalate to a senior scorer or physician. - **Operational handover.** A written runbook for your team: where data lives, how to back it up and test a restore, how updates are handled, and who to contact. ## Deployment, step by step 1. **Consultation.** We learn how your lab scores today, who reviews, and what you want from local, software-assisted staging. 2. **Environment assessment.** We check workstations, recording exports and IT requirements, and agree what must be in place before any study is processed. 3. **Configuration and pilot.** Installation and configuration, then a pilot on agreed recordings so your team can judge fit before wider use. 4. **Training and handover.** Technician training, guided review and a documented operational handover. 5. **Agreed support.** Ongoing support on the terms written into your deployment agreement. Platforms: a standalone macOS build for Apple silicon (macOS 14 or later) has been built and tested. A Windows x64 package has been prepared but not yet verified on Windows hardware. ## Frequently asked questions ### Who is InEpoch for? Sleep laboratories, hospital sleep departments, clinical operations leads, sleep technologists and sleep researchers who stage polysomnography or EEG recordings in 30-second epochs and want local, software-assisted staging with structured human review. ### Where is study data processed? On your workstation. The application stores recordings and results in a local data folder and runs staging on the CPU; no cloud upload is required for analysis. Access, disk encryption, backups and any support access are configured with your organization. This website is separate from the software, and its inquiry form asks only for contact details. ### What does installation involve? An environment assessment, installation on the agreed workstations, a check that your recordings import correctly, and configuration of the review workflow. A macOS build for Apple silicon (macOS 14 or later) has been built and tested. A Windows x64 package has been prepared and still needs to be verified on Windows hardware, so platform fit is confirmed during the assessment. ### How does training work? Hands-on sessions with your technologists and reviewers on the workstation, using authorized or public sample recordings, followed by guided review and a written operational handover. Format and duration are agreed with you; there is no fixed curriculum. ### What do reviewers do? Reviewers check the proposed stage for each epoch against the signals, confirm or correct it, and add notes where useful. The YASA documentation recommends that a trained scorer check its predictions, especially low-confidence and N1 epochs, and the review queues are designed around that. ### Does InEpoch diagnose sleep disorders? No. The software proposes sleep stages for professional review; it does not make diagnoses or replace a qualified professional. It does not score respiratory events, arousals or limb movements, and it does not compute apnea indices. No regulatory clearance, certification or independent clinical validation is claimed. ### Is it ready for clinical use in our lab? We start with a deployment assessment and a pilot. Before any clinical use we review with you the licensing of the pretrained models, the regulatory position in your jurisdiction, and local validation against your own scorers. ### How do we discuss the scope of a deployment? Use the form below to tell us about your lab, your recording setup and what you would like to change. We'll reply to arrange a conversation or a demonstration. ## Guides - [What is sleep staging?](https://inepoch.com/learn/what-is-sleep-staging): Sleep staging splits a sleep recording into 30-second epochs labelled Wake, N1, N2, N3 or REM. How it works, which signals are used, what a hypnogram shows. - [AI-assisted sleep scoring: how it works and where people fit](https://inepoch.com/learn/ai-assisted-sleep-scoring): How AI sleep-staging software proposes a stage for every 30-second epoch, where algorithms go wrong, how autoscoring is certified, and what reviewers do. - [YASA sleep staging explained: inputs, accuracy and limits](https://inepoch.com/learn/yasa-sleep-staging): What the open-source YASA sleep-staging algorithm needs, how it was trained and tested, what its authors report about accuracy, and its known limitations. - [Local vs cloud sleep scoring: where study data is processed](https://inepoch.com/learn/local-vs-cloud-sleep-scoring): Local sleep-scoring software analyses recordings on your workstation; cloud services use a vendor's servers. Compare data flow, IT duties and what to ask. - [How to evaluate sleep staging software: a checklist for sleep labs](https://inepoch.com/learn/sleep-staging-software-checklist): A checklist for sleep labs comparing sleep-staging or autoscoring software: validation, certification, review tools, data handling, formats and support. - [Sleep staging glossary](https://inepoch.com/learn/sleep-staging-glossary): Plain-language definitions of sleep staging terms: epoch, hypnogram, PSG, EEG, EOG, EMG, EDF, N1, N2, N3, REM, spindles, K-complexes, TST, WASO and more. ## Contact Discuss a deployment through the inquiry form at [inepoch.com/#contact](https://inepoch.com/#contact). ## Sources - [AASM Scoring Manual](https://aasm.org/clinical-resources/scoring-manual/) - [AASM position statement on AI in sleep medicine (2020)](https://aasm.org/advocacy/position-statements/artificial-intelligence-sleep-medicine-position-statement/) - [Vallat & Walker, eLife (2021)](https://doi.org/10.7554/eLife.70092) - [YASA documentation: FAQ](https://yasa-sleep.org/faq.html) - [YASA on GitHub (BSD-3-Clause)](https://github.com/raphaelvallat/yasa)